#
Robust Weighted Least Squares Support Vector Regression algorithm to estimate the nanofluid thermal properties of water/graphene Oxide-Silicon carbide mixture

Robust Weighted Least Squares Support Vector Regression algorithm to estimate the nanofluid thermal properties of water/graphene Oxide-Silicon carbide mixture

A new optimization/statistical approach of “Robust Weighted Least Squares Support Vector Regression” algorithm (RWLS-SVR) is provided for the first time. The experimental achieved amounts of the thermal conductivity for a new hybrid nanofluid of water/Graphene Oxide–Silicon Carbide, are examined at different values of temperature and nanoparticles volume fraction. A Support Vector Regression is a supervised learning regression algorithm based on the Support Vector Machine methodology. However in the Least Squares Support Vector Machine method, the inequality constraints are converted to equality constraints and the sum squared error function is employed. Moreover a LS-SVR is applied to the problem in order to calculate the error variables. Afterwards, the weights computed based on the error variables are applied to the optimization problem in order to reduce the effects of outliers on the final results. As a result, the WLS-SVR method does not significantly increase the computational burden, but it provides sparseness and robustness.